this may come as a shock to every one here but I have successfully designed ,tested and flown an 6DOF IMU based AP including extended Kalman filter ,centripetal acceleration compensation and Z-axis gyro biasing through GPS . i have posted in my netburner blog.and no i havn't been paid a dime .
The bottom line appears to be:
For fixed-wing platforms with neutral roll stability,
many ppl did roll stabilisation using yaw gyro or GPS turnrate to roll formula. Apparently nobody made a system that would truly work both stationary and in coordinated turn during 20s (high-class direct feed to rollangle estimate from roll gyro), albeit many claim this is easy, I claim this is invariably spin condition unless using 400USD ADXL IMU or similar.
IMU copter pilots are different story as they don't move in coordinated turn for longer than a few s. On the other hand they have troubles centering the yaw gyro, hence the magnetometers come into play and GPS course is invalid for mostly stationary flights. This limits current amateur state-of-the-art IMU based neutral stability (fixed wing like flying wing) planes for flying at speeds significantly higher than windspeed in order to be sure to get reliable GPS course (since I cannot believe you can call cheap-yaw-gyro based system a reliable thing without gyro centering). Self-stable IMU planes based on our achievements do not seem to have this limitation for long-term roll stability, but you still need GPS course for navigation.
I have been flying IMU based systems for 5 years, using boards that SparkFun and I designed and firmware that I write. There are two generations of boards. My first board has two gyros, two accelerometers, and a GPS. It uses simple controls to stabilize a Gentle Lady and provides a return to launch function. Pitch is based on direct accelerometer and gyro feedback. Yaw is based on GPS and gyro. It has reliably performed for me for over a hundred flights, saved me from losing my plane a few times.
My new board has 3 gyros and 3 accelerometers. Paul Bizard and I recently developed an algorithm that integrates the nonlinear differential equations that describe the kinematics of rotation, which we call direction-cosine-matrix (DCM). Its performance in a roll-pitch-yaw demo is astounding. UFO-MAN took a video.
A port of my previous 2DoF firmware to the new board did not work out, because of the dynamics of the EM406, so I abandoned it and went straight to the DCM version, which is almost done, and in the final stages of flight testing. Its been up a dozen times, works well. It really does not excercise the full capabilities of the board or the DCM algorithm, but it is a good stepping stone for me to my next goals. As soon as its done and released, I am going to move on to my next two projects, which are going to be to stabilize Matt Chang's Harrier Jump Jet, and to add a stabilization co-pilot to an aerobatic plane that I have trouble flying otherwise.
I have not achieved a complete roll-leveling loop yet, but based on my roll-pitch-yaw demo, it looks like it is going to be rather easy to do. I will have accurate estimates of the 9 direction cosines describing the orientation of the plane, with 0.020 second transient response, and zero drift.
Well, the original question was about amateur roll leveling feedback. Obviously amateurs are getting roll leveling feedback from IMU's in vehicles that leave the ground.
If the issue is whether that roll leveling feedback is being sold in stores or on a fixed wing, that changes matters. Personally define professional as anything that makes money, is funded by a professional operation, or is tax deductable.
It's really hard to make money. You're not amateur anymore if you're doing that. You're pro.
Ardupilot is professional. That feeds Sparkfun.
Curt Olson is a pro. He's funded by the University of Minnesota.
Jack Crossfire is an amateur.
And, for what it's worth, I've flown the my Lego Mindstorms autopilot with a crappy PID-based IMU. I can't say that it flew well, but it did stay up there in a wobbly sort of way ;-)
I've been flying my "MicroGear" based system for about a year and a half now. This is an IMU based system (not IR sensor based.) Recently I did a 35 minute flight of our flying wing "maranized" UAV entirely over water ... launched from a small boat and recovered in the water.
My website / mini-blog is kind of out of date, but here's a quick update:
The original code was designed around the Xbow MNAV IMU which is now defunct and unavailable. So I am in the process of evaluating two possible/partial replacements. 1. The Spark Fun 6DOF v4 IMU, and a new Analog Devices 23mm cube IMU. I have the spark fun working on my desk and so far so good. We don't have the analog devices IMU here yet, but it's in the order pipeline. Both of these have 3 axis gyro, 3 axis accelerometer, and 3 axis magnetometer.
I've also added generic support for "gpsd" which is a linux based gps driver that can speak just about every gps protocol out there (well a lot of them anyway), supports serial and usb gps's, and presents the results in a consistent way to MicroGear. This is a convenient/easy way to support a boatload of different GPS's.
In addition I'm working on adapting a much more sophisticated (16 state) kalman filter algorithm developed by an aerospace engineering professor at the University of Minnesota. That process is well along, and early results are very promising, both in terms of algorithm speed and correctness. This more sophisticated algorithm produces a *much* more accurate location and attitude estimate than we could ever get out of the Xbow open-source ahrs/nav algorithm.
Right now I continue to develop for linux/gumstix, but we are also looking at the MPC-5200 running eCos, and the code should compile run on just about any modern linux/unix/posix based embedded system.
All these new improvements will be released as open-source, but currently we are doing quite a bit of testing and code massaging in house and don't plan to do any public release until we get all the pieces working together. (Previous versions are still available for download.)
I get asked questions about ground station code, and unfortunately, our ground station software is developed under contract and not open-source. Hopefully we can figure out a way to make it available in some reasonable way.
I'm really looking forward to get these pieces all put together in an airplane this spring. The performance and accuracy should be outstanding and should leap frog many of the available commercial systems. And the onboard autopilot code will be all open source, and the communication protocol with the operator station will be open.
We still have some issues to work through, but I'm very excited!
According to Chris's interview, the Paparazzi code actually flying is neither amateur nor IMU guided. It was started by the École Nationale de l'Aviation Civile & it uses IR.
Probably for DIYDrones, but all those autonomous tri rotor, quad rotor, T-Rex, & Corona videos on Goo Tube were amateur. Millions of amateurs never bother documenting it on the internet. Thousands document it but never get discovered by the Goog. Of those, even less ever upload pictures. Of all the IMU's flying, only a really tiny fraction were built by pros. There isn't enough money in the industry for that many pros.
Replies
For fixed-wing platforms with neutral roll stability,
many ppl did roll stabilisation using yaw gyro or GPS turnrate to roll formula. Apparently nobody made a system that would truly work both stationary and in coordinated turn during 20s (high-class direct feed to rollangle estimate from roll gyro), albeit many claim this is easy, I claim this is invariably spin condition unless using 400USD ADXL IMU or similar.
IMU copter pilots are different story as they don't move in coordinated turn for longer than a few s. On the other hand they have troubles centering the yaw gyro, hence the magnetometers come into play and GPS course is invalid for mostly stationary flights.
This limits current amateur state-of-the-art IMU based neutral stability (fixed wing like flying wing) planes for flying at speeds significantly higher than windspeed in order to be sure to get reliable GPS course (since I cannot believe you can call cheap-yaw-gyro based system a reliable thing without gyro centering). Self-stable IMU planes based on our achievements do not seem to have this limitation for long-term roll stability, but you still need GPS course for navigation.
My new board has 3 gyros and 3 accelerometers. Paul Bizard and I recently developed an algorithm that integrates the nonlinear differential equations that describe the kinematics of rotation, which we call direction-cosine-matrix (DCM). Its performance in a roll-pitch-yaw demo is astounding. UFO-MAN took a video.
A port of my previous 2DoF firmware to the new board did not work out, because of the dynamics of the EM406, so I abandoned it and went straight to the DCM version, which is almost done, and in the final stages of flight testing. Its been up a dozen times, works well. It really does not excercise the full capabilities of the board or the DCM algorithm, but it is a good stepping stone for me to my next goals. As soon as its done and released, I am going to move on to my next two projects, which are going to be to stabilize Matt Chang's Harrier Jump Jet, and to add a stabilization co-pilot to an aerobatic plane that I have trouble flying otherwise.
I have not achieved a complete roll-leveling loop yet, but based on my roll-pitch-yaw demo, it looks like it is going to be rather easy to do. I will have accurate estimates of the 9 direction cosines describing the orientation of the plane, with 0.020 second transient response, and zero drift.
Bill Premerlani
If the issue is whether that roll leveling feedback is being sold in stores or on a fixed wing, that changes matters. Personally define professional as anything that makes money, is funded by a professional operation, or is tax deductable.
It's really hard to make money. You're not amateur anymore if you're doing that. You're pro.
Ardupilot is professional. That feeds Sparkfun.
Curt Olson is a pro. He's funded by the University of Minnesota.
Jack Crossfire is an amateur.
http://baron.flightgear.org/~curt/UAS/MicroGear1/
My website / mini-blog is kind of out of date, but here's a quick update:
The original code was designed around the Xbow MNAV IMU which is now defunct and unavailable. So I am in the process of evaluating two possible/partial replacements. 1. The Spark Fun 6DOF v4 IMU, and a new Analog Devices 23mm cube IMU. I have the spark fun working on my desk and so far so good. We don't have the analog devices IMU here yet, but it's in the order pipeline. Both of these have 3 axis gyro, 3 axis accelerometer, and 3 axis magnetometer.
I've also added generic support for "gpsd" which is a linux based gps driver that can speak just about every gps protocol out there (well a lot of them anyway), supports serial and usb gps's, and presents the results in a consistent way to MicroGear. This is a convenient/easy way to support a boatload of different GPS's.
In addition I'm working on adapting a much more sophisticated (16 state) kalman filter algorithm developed by an aerospace engineering professor at the University of Minnesota. That process is well along, and early results are very promising, both in terms of algorithm speed and correctness. This more sophisticated algorithm produces a *much* more accurate location and attitude estimate than we could ever get out of the Xbow open-source ahrs/nav algorithm.
Right now I continue to develop for linux/gumstix, but we are also looking at the MPC-5200 running eCos, and the code should compile run on just about any modern linux/unix/posix based embedded system.
All these new improvements will be released as open-source, but currently we are doing quite a bit of testing and code massaging in house and don't plan to do any public release until we get all the pieces working together. (Previous versions are still available for download.)
I get asked questions about ground station code, and unfortunately, our ground station software is developed under contract and not open-source. Hopefully we can figure out a way to make it available in some reasonable way.
I'm really looking forward to get these pieces all put together in an airplane this spring. The performance and accuracy should be outstanding and should leap frog many of the available commercial systems. And the onboard autopilot code will be all open source, and the communication protocol with the operator station will be open.
We still have some issues to work through, but I'm very excited!
Curt.
http://diydrones.com/profiles/blog/show?id=705844%3ABlogPost%3A29982
However the amateur IMU's today are improving fast. They're being flown without accelerometers & compasses. They're using GPS to correct heading.
-
1
-
2
of 2 Next